# What people change after AI drafts a course

> After an AI course draft, people add content (17% of requests), images (14%) and quizzes (10%). Only 1.5% ask for a shorter course.

**Author:** Parshad Desai  
**Published:** 2026-11-09  
**Categories:** Instructional Designers, Course Creators

---

The first draft of an AI course is rarely the last. Three in four people keep talking to the AI before the course is even generated.

I looked at what people on Coursebox ask the AI to change once a course exists, using a sample of 800 editor requests from 2,260 people, and at what they add to the course by hand. **People mostly add: new content, images and quizzes lead the list. Requests to shorten (1.5%) or simplify (2%) a course are rare.**

> From the Coursebox Research Report: [The State of AI Course Creation 2026](https://www.coursebox.ai/research/state-of-ai-course-creation)

## How many prompts do people send before the first draft?

76% of people send more than one prompt before their course is generated, according to Coursebox's data on 37,284 people (May to November 2026). The median is two prompts, and a quarter send three or more.

_Prompts sent before the course is generated_

| Category | Value |
| --- | --- |
| One prompt | 24% |
| Two or more prompts | 76% |

Base: 37,284 people, May to Nov 2026

The follow-up prompts usually fill in what the first brief left out: the audience, the length, or a document to work from. The conversation has become part of the brief.

## What do people ask the AI to change after the first draft?

After the first draft, the most common request is to add new content (17% of requests), followed by images (14%), a quiz (10%) and examples or scenarios (8%), according to Coursebox's analysis of 800 editor AI requests from 2,260 people (April to November 2026). 12% of requests are questions, such as how to do something in the course.

_What people ask the AI to change_

| Category | Value |
| --- | --- |
| Add new content | 17% |
| Add images | 14% |
| Ask a question | 12% |
| Add a quiz | 9.5% |
| Rewrite a section | 9% |
| Change format or layout | 9% |
| Add examples or scenarios | 8% |
| Add interactive blocks | 6% |
| Restructure | 6% |
| Fix errors | 4% |

Base: 800 editor AI requests from 2,260 people (max 3 per person), Apr to Nov 2026. A request can include several changes

63% of requests are specific, naming the section, the number of questions or the change wanted. The other 37% are vague, such as "make it better", and those leave the AI to guess.

## How often do people ask for a shorter or simpler course?

Rarely: 1.5% of editor requests ask the AI to shorten a course and 2% to simplify it, against 17% that ask to add content. Fixing errors accounts for 4%, and aligning the course to a source document for 3.5%.

_Requests that remove or correct, against requests that add_

| Category | Value |
| --- | --- |
| Add new content | 17% |
| Fix errors | 4% |
| Align to the source document | 3.5% |
| Simplify | 2% |
| Shorten | 1.5% |

Base: 800 editor AI requests from 2,260 people (max 3 per person), Apr to Nov 2026. A request can include several changes

Adding is the natural instinct once a draft exists, because gaps are easier to see than excess. For learners, a shorter, plainer module is often the bigger improvement, which makes it worth checking for before anything else is added.

## What do people add to the course themselves?

Video is the content type people add most once the draft is done: 48% of editors add a video block, 32% an image and 28% a quiz, according to Coursebox's data on 2,860 editors (August to October 2026). Assignments (16%) and flip cards (14%) are used by a smaller group.

_Share of editors who added each type of content_

| Category | Value |
| --- | --- |
| Video | 48% |
| Image | 32% |
| Quiz | 28% |
| Notes and text | 23% |
| Content block | 19% |
| Assignment | 16% |
| Flip cards | 14% |

Base: 2,860 editors, Aug to Oct 2026

The pattern suggests people treat the AI draft as the structure and text, then add the media and practice themselves.

## How should you review an AI course draft?

Review the draft against what learners need before adding to it. This is the prompt I use for that review; paste it into the AI editor once the course is generated:

**Prompt: review an AI course draft**

Paste into the AI editor after the first draft is generated, before adding more content.

```
Review this course for [AUDIENCE], who need to be able to [OUTCOME] after [LENGTH].

For each module:
1. Flag anything that does not serve that outcome and suggest cutting it.
2. Flag any page longer than [MAX_WORDS] words and suggest a shorter version.
3. Check that it ends with practice: a quiz question, scenario or task that tests the outcome.
4. List any facts I should check against [SOURCE].

Show me the suggested changes before applying them.
```

Replace these:
- `[AUDIENCE]`: Who the course is for (for example, new customer service staff)
- `[OUTCOME]`: What learners should be able to do (for example, handle a refund request using the company policy)
- `[LENGTH]`: How long learners have (for example, a 20-minute course)
- `[MAX_WORDS]`: Longest page you want (for example, 250)
- `[SOURCE]`: The document the facts must match (for example, refunds policy)

Check before you use it:
- Every module ends with practice
- Nothing is cut without your approval
- Facts to check are listed against your source

## How I measured this

The data is anonymised and aggregated. No names, email addresses or company details were used in the analysis, no prompt text is published, and any group smaller than 30 is left out. Prompt counts cover 137,126 course-creation prompts from 37,284 people on Coursebox between 14 May and 9 November 2026. Editor requests are an AI-classified random sample of 800 requests from 2,260 people (at most three per person) between April and November 2026; a request can include several changes. Content blocks come from product events from 2,860 editors between 1 August and 31 October 2026. Bots, staff and test accounts are removed throughout. The full methodology is in the report.